does read order influence bwa mapping?

[jfeicheng]

I'm doing alignment with bwa (0.7.5a-r405).
But I found that the same sequence in different position may influence the mapping result.

As you may noticed, the read1 and read3 are exactly the same, but the mapping result is different.

read1 0 Chr3 209 0 50M * 0 0 CAAAGGTGCGGCCCGAGGTAAAGGGTTTAACGCTTCATATTTTATCACAT HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH XT:A:R NM:i:0 X0:i:2 X1:i:1 XM:i:0 XO:i:0 XG:i:0 MD:Z:50 XA:Z:Chr11,-26081473,50M,1;
read2 0 Chr3 208 0 50M * 0 0 ACAAAGGTGCGGCCCGAGGTAAAGGGTTTAACGCTTCATATTTTATCACA HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH XT:A:R NM:i:0 X0:i:2 X1:i:1 XM:i:0 XO:i:0 XG:i:0 MD:Z:50 XA:Z:Chr3,+4840,50M,0;Chr11,-26081474,50M,1;
read3 4 * 0 0 * * 0 0 CAAAGGTGCGGCCCGAGGTAAAGGGTTTAACGCTTCATATTTTATCACAT HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH

Does anyone know what's happening here?

the command line:
input=$1
bwa aln genome.fa $input >${input/.fq/}.sai
bwa samse -f ${input/.fq/}.sam genome.fa ${input/.fq/}.sai $input

[dariober]

The mapping quality is 0 meaning that there are multiple and equally probable alignments. In this case bwa picks one location at random and this could explain why different reads with the same sequence are reported at different positions.